• 검색 결과가 없습니다.

Regional Factors Associated with Participation in the National Health Screening Program: A Multilevel Analysis Using National Data

N/A
N/A
Protected

Academic year: 2021

Share "Regional Factors Associated with Participation in the National Health Screening Program: A Multilevel Analysis Using National Data"

Copied!
9
0
0

로드 중.... (전체 텍스트 보기)

전체 글

(1)

Regional Factors Associated with Participation in the National Health Screening Program: A Multilevel Analysis Using National Data

High participation rates are important for maximizing the effects of a health screening program. Previous studies have suggested that individual or regional characteristics have effects on health behaviors. In this study, we investigated the determinants of participation in the National Screening Program for Transitional Ages by simultaneously analyzing individual and area-level factors by multilevel analysis. A total of 1,081,216 subjects, aged 40 and 66 yr and nested in 254 areas, were included. There was a significant variation in participation rates across the areas even after adjusting for individual and area-level variables. Among the individual-level variables, increasing age, sex, higher income, and mild disability grade were associated with a higher participation rate. In urban areas, the 40-yr-old group had higher participation rates than the 66-yr-old group. Deprived areas had significantly high participation rates for both age groups. The number of screening centers per 1,000 inhabitants had no significant effect on participation in the screening program. In conclusion, regional characteristics are associated with participation rates independent of personal features and regional factors have differential effects with respect to age. A multi-dimensional approach is recommended to promote participation in health screening programs.

Key Words: Health Screening; Participation Rate; Multilevel Analysis; Regional Factors;

National Screening Program for Transitional Ages Hyung-Kook Yang,1 Dong-Wook Shin,1

Seung-Sik Hwang,2 Juwhan Oh,3 and Be-Long Cho1

1Department of Family Medicine, Seoul National University Hospital, Seoul; 2Department of Social and Preventive Medicine, Inha University School of Medicine, Incheon; 3Institute of Health Policy and Management, Medical Research Center, Seoul National University, Seoul, Korea

Received: 9 July 2012 Accepted: 11 January 2013 Address for Correspondence:

Be-Long Cho, MD

Department of Family Medicine, Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul 110-744, Korea

Tel: +82.2-2072-2195, Fax: +82.2-766-3276 E-mail: belong@snu.ac.kr

This study was funded by the grant 2010E3303200 from the Korea Centers for Disease Control and Prevention.

http://dx.doi.org/10.3346/jkms.2013.28.3.348 • J Korean Med Sci 2013; 28: 348-356

INTRODUCTION

Health screening programs aim primarily to reduce the mortal- ity of target diseases by early detection and effective treatment (1). High participation rates should be ensured to maximize the benefits of such screening programs. It is known that individual factors such as age, sex, disability (2, 3), and socioeconomic sta- tus (4-6) affect the rate of participation. Moreover, psychosocial factors such as knowledge and beliefs about a program, attitude towards a screening program, and cultural background (7-9) are associated with participation. However, previous studies have mainly focused on individual characteristics and largely ignored regional characteristics.

Understanding the effects of regional characteristics is im- portant because our health and behavior are affected by the environment and features of the region we inhabit (10). Addi- tionally, some studies have reported that regional socioeconom- ic factors such as area income, unemployment, and education levels significantly affect the participation rate in health screen- ing programs independent of individual characteristics (11-13).

However, those studies were conducted as parts of cancer screen-

ing programs (8, 14-17) or involved study populations in limit- ed areas (18, 19). Furthermore, individual and area-level vari- ables were separately analyzed using single-level multivariable regression (12, 20).

The Republic of Korea Government has actively carried out health screening programs in an effort to detect diseases early and to improve public health. Since their introduction in the 1980s, the coverage of periodic national health screening pro- grams has been expanding. In 2007, a new national health screen- ing program, the National Screening Program for Transitional Ages, was added to the existing program (21). It covers those who are 40 and 66 yr old, including Medicaid recipients who were excluded from the national general health screening pro- gram.

In this study, using the whole data of the National Screening Program for Transitional Ages, we simultaneously analyzed in- dividual and area-level factors by multilevel analysis to find the determinants of participation in the national screening program.

Multilevel analysis allows the examination of area and individ- ual-level factors simultaneously. Most studies in Korea have ap- proached individual or area-level factors separately or controlled

(2)

different level factors at the same level. This may have led to over/

underestimation of the effect of different level factors. Multilev- el analysis can examine both inter-individual and intergroup variation and allows researchers to deal simultaneously with individuals at a micro-level and areas at a macro-level.

Although previous studies have suggested various individual and regional determinants of participation in health screening programs, we selected individual and area-level factors by con- sidering practical and political implications. Eventually, we se- lected 3 domains based on regional characteristics: level of ur- banization, regional deprivation, and medical supply. We hy- pothesized that more urbanized and less deprived areas would have higher participation rates, and that the number of screen- ing centers would exhibit a positive relation with participation.

MATERIALS AND METHODS Data sources

Our study is based on the data of the National Screening Pro- gram for Transitional Ages in 2008 and 2009 obtained from the National Health Insurance Corporation (NHIC). In Korea, the National Health Insurance provides mandatory universal health insurance to nearly all Koreans (96.9%); others are covered by a public assistance plan (i.e., Medicaid). As the Ministry of Health and Welfare entrusts the handling of claims to the NHIC, the data of Medicaid enrollees are also managed by the NHIC. We excluded the insured employee group from our study popula- tion because both employers and employees may incur penal- ties if they do not undergo health checkups. As the NHIC pro- vided data without identification codes for the dependents of the employee insured group at the beginning of our study, we could not analyze these dependents in our study.

Among the data of 1,081,249 potentially eligible subjects in 254 municipal districts, the personal health registry data of 1,081,216 were used after excluding subjects who had duplicate or missed identification numbers. Eligible area-level data were collected from various sources. We merged the area and individual-level data using the local address codes of the municipal districts (Fig. 1).

Individual-level variables

Individual-level variables were derived from the insurance reg- istry data of the NHIC. The residential address codes of the study population were all deleted and replaced with simple numbers to ensure their privacy. We selected the final individual vari- ables according to a review of previous studies and ease of in- terpretation. The variables are as follows: 1) year (2008 vs 2009), 2) age (40 vs 66 yr old), 3) sex (male vs female), 4) health insur- ance type (self-employed insured vs Medicaid recipients), 5) health insurance premium, and 6) disability grade according to the legal classification of severity according to Article 2 of the Welfare of Disabled Persons Act (22).

Health insurance premiums are determined according to monthly income levels in Korea; we used health insurance pre- mium as an indicator of individual income level. We divided health insurance premium into quintiles from group 1 (lowest) to group 5 (highest) according to the monthly health insurance premium amount. The Medicaid recipient group, who does not pay at all, was designated as group 0. Disability was graded from 1 (severe) to 6 (mild) according to the legal category of disability severity. Subjects without any disability were categorized as the non-disabled group.

Area-level variables

Initially, we collected 199 area-level variables from various sourc- es, such as the 2005 Population Census of the Korea National Statistical Office; 2005–2006 Benefit Recipient data; 2008 Kore- an Medical Association’s membership survey; 2008 land regis- tration statistics of the Ministry of Land, Transport, and Mari- time Affairs; and the 2010 National Health Insurance Corpora- tion. Among 199 potential variables, we chose 3 area-level vari- ables based on statistical significance from univariate analyses and health policy implications: 1) the proportion of agriculture, forestry, and fishery workers (%) as an inverse index of the ur- banization of the region; 2) the Composite Deprivation Index (CDI) score, which represents the level of deprivation in an area (23); and 3) the number of screening centers per 1,000 inhabit- ants, which indicates the level of medical supply. All variables except CDI score were based on data from 2007.

We used the CDI as the parameter for regional deprivation.

The CDI is calculated according to 5 domains: 1) unemploy- ment, 2) poverty, 3) housing, 4) labor, and 5) social network.

The proxy variables of the 5 domains are as follows: 1) the pro- Fig. 1. Study framework.

Study population 1,081,216 subjects nested in

254 municipal districts

Indentification number duplication & omission (33 subjects excluded)

1,081,249 subjects Individual level variables

Subjects of the National Screening Program for transitional ages in 2008 and 2009

National Health Insurance registry data

Year, age, sex, health insurance premium, Disability Grade

254 municipal districts Area level variables

The proportion of agriculture, forestry and fishery workers (%)

Composite Deprivation Index (CDI) score

Numbers of the screening centers per 1,000 inhabitants

Address code

1,081,216 subjects 254 municipal districts

(3)

portion of unemployed males, 2) the percentage of recipients receiving National Basic Livelihood Security Act benefits, 3) the proportion of households under the minimum housing stan- dard, 4) the proportion of people with low social class, and 5) the proportion of single-parent households (Table 1). Previous studies have reported that the CDI was more suitable in Korea than the Townsend or Carstairs indices, which were developed in England and widely used as indicators for regional depriva- tion (23, 24).

Statistical analysis

We divided the study population into 2 groups according to age:

subjects were 40 yr and 66 yr old. Multilevel analysis was per- formed on 1,081,216 individuals (level 1) nested within 254 mu-

nicipal districts (level 2). A 2-level logistic regression model us- ing the Markov Chain Monte Carlo method was constructed to estimate the average relationships between participation in the national health screening program and individual-level variables across all regions (i.e., individual fixed parameters) and between the variation between municipal districts (i.e., regional random variance), and the effects of area-level variables on participa- tion in the screening program (i.e., regional fixed parameters).

We performed multilevel logistic regression analyses adjust- ing for both individual and area-level variables as fixed effects and allowed for heterogeneity between areas on a step-by-step basis. First, a null model with no explanatory variable was ana- lyzed for the existence of variance between areas (null model).

Second, to calculate the proportion of variance related to the

Table 1. Description on the Composite Deprivation Index

Domain Indicator Definition Data Source

Unemployment Male unemployment proportion Percentage of male persons aged 15-64 yr

who are economically active and seeking job 2005 Population and Housing Census Poverty Recipients of National Basic Livelihood

Security Act benefits

Percentage of persons who receive National Basic Livelihood Security Act benefits

2005-2006 mean-tested benefit recipients data Housing Households under the minimum

housing standard

Percentage of households under the minimum housing standard (bedroom standards, housing facilities standard, floor space standard, and the separate bedroom principle)

2005 Population and Housing Census

Labor Low social class Percentage of all persons in households with a head of the household who is engaged in elementary occupations

2005 Population and Housing Census

Social network Single-parent household Percentage of all persons in a single parent household where

the parent is under 60 yr old 2005 Population and Housing Census

Table 2. Baseline characteristics of the study population

Parameters

40-yr-old group 66-yr-old group

Non-participants No. (%) 517,185 (67.6)

Participants No. (%)

248,479 (32.5) P value* Non-participants No. (%) 144,833 (47.4)

Participants No. (%)

160,718 (52.6) P value*

Year 2008

2009 284,128 (69.8)

238,702 (65.5) 122,962 (30.2) 125,895 (34.5)

< 0.01

89,832 (49.6)

58,385 (45.4) 91,182 (50.4) 70,130 (54.6)

< 0.01

Sex Male

Female 282,104 (74.6)

240,726 (61.1) 95,839 (25.4) 153,018 (38.9)

< 0.01

69,002 (49.6)

79,215 (46.5) 70,116 (50.4) 91,196 (53.5)

< 0.01

Health Insurance Type Medicaid

The Self employ Insured 28,786 (67.5)

494,044 (67.8) 13,861 (32.5) 234,996 (32.2)

< 0.01

27,478 (59.2)

120,739 (45.9) 18,936 (40.8) 142,376 (54.1)

< 0.01

Insurance Premium Medicaid 1st quintile (lowest) 2nd quintile 3rd quintile 4th quintile 5th quintile (highest)

28,784 (67.5) 109,977 (75.6) 111,516 (73.0) 101,104 (66.8) 90,339 (61.7) 75,465 (59.4)

13,851 (32.5) 35,472 (24.4) 41,204 (27.0) 50,325 (33.2) 56,087 (38.3) 51,540 (40.6)

< 0.01

27,477 (59.2) 27,584 (52.4) 19,355 (45.9) 19,827 (43.7) 21,442 (43.1) 29,148 (42.1)

18,923 (40.8) 25,033 (47.6) 22,814 (54.1) 25,542 (56.3) 28,341 (56.9) 40,065 (57.9)

< 0.01

Disability Grade Not disabled Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Grade 6

480,048 (67.4) 3,464 (74.0) 5,969 (72.0) 5,291 (68.9) 2,636 (69.1) 3,502 (69.4) 5,683 (68.9)

232,580 (32.6) 1,220 (26.1) 2,321 (28.0) 2,384 (31.1) 1,177 (30.9) 1,547 (30.6) 2,563 (31.1)

< 0.01

108,947 (45.3) 2,910 (77.7) 4,750 (66.8) 4,223 (55.1) 3,432 (45.9) 4,192 (43.3) 4,044 (43.5)

131,518 (54.7) 836 (22.3) 2,366 (33.3) 3,438 (44.9) 4,051 (54.1) 5,499 (56.7) 5255 (56.5)

< 0.01

*Results of P value were performed by Chi-square test.

(4)

region remaining after adjusting for individual-level variables, individual-level variables were added into the null model (Mod- el 1). Finally, individual and area-level variables were simulta- neously introduced into the model (Model 2). The area-level random effect of the intercept was assumed to have a normal distribution. We calculated the proportion of variance related to the region (intraclass correlation: ICC) as approximately: re- gional variance/(regional variance + π2/3). In addition, we test- ed area-level factors using a random slope model, and tested interactions between individual-level and area-level factors.

All statistical analyses were performed using STATA 12.0 (STATA Corp, Houston, TX, USA) and MLwiN 2.25 (University of Bristol, UK). Statistical significance was defined as a 2-sided P value less than 0.05.

Ethics statement

This study protocol was approved by the institutional review board of the Seoul National University Hospital (IRB No. H 1109- 031-377). Informed consent was waived.

RESULTS

Baseline characteristics of the participants

Table 2 lists the baseline characteristics of the study population.

Of the 1,081,216 subjects, 410,168 (37.9%) participated in the

screening program. The participation rate of the 40-yr-old group was lower than that of the 66-yr-old group (30.2% vs 50.4% in 2008; 34.5% vs 54.6% in 2009). More women participated in the screening program. The sex gap was more prominent in the 40-yr-old group than in the 66-yr-old group (13.5% vs 2.6%). In- come level (health insurance premium) was generally positive- ly related to the participation rate in both age groups, except in the first and second income quintiles in the 40-yr-old group.

Severity of disability was negatively associated with participa- tion rate.

Multilevel analysis in the 40-yr-old group

Table 3 shows the results of multilevel analysis, which included individual and area-level variables in a hierarchical manner.

There was significant variance with respect to participation rates across the area in the null model (log-odds [β], 0.04; standard error [SE], 0.004) (Fig. 2). When the null model was adjusted with individual-level variables (Model 1), the regional variance decreased (β, 0.036; SE, 0.004). All individual-level variables were significantly associated with participation rate. After the null model was simultaneously adjusted with individual and area-level variables (Model 2), the regional variance decreased further (β, 0.031; SE, 0.003). However, some unexplained vari- ance across areas was still observed.

There were lower participation rates in the first and second

Table 3. Multilevel analysis results of the 40-yr-old group

Response Null Model Model 1 Model 2

β (S.E) β (S.E.) β (S.E.)

Fixed Part

Constant -0.785 (0.011) -1.263 (0.016) -1.331 (0.038)

2009 (vs 2008) 0.200 (0.005) 0.189 (0.006)

Female (vs male) 0.610 (0.005) 0.605 (0.005)

Health Insurance Premium Medicaid Recipient 1st quintile 2nd quintile 3rd quintile 4th quintile 5th quintile

Reference -0.421 (0.014) -0.213 (0.012) 0.041 (0.012) 0.303 (0.012) 0.422 (0.012)

Reference -0.445 (0.018) -0.228 (0.017) 0.024 (0.017) 0.282 (0.017) 0.399 (0.017) Disability Grade

No-disabled 1st grade 2nd grade 3rd grade 4th grade 5th grade 6th grade

Reference -0.161 (0.035) -0.080 (0.026) 0.143 (0.025) 0.184 (0.037) 0.179 (0.033) 0.238 (0.024)

Reference -0.135 (0.037) -0.083 (0.028) 0.142 (0.028) 0.194 (0.038) 0.192 (0.034) 0.250 (0.027)

The proportion of agriculture, forestry and fishery workers (%) -0.005 (0.002)

Composite Deprivation Index Score 0.001 (< 0.0001)

Numbers of the screening center per 1,000 inhabitants 0.493 (0.404)

Random Part : Area Level

Variance (S.E) 0.04 (0.004) 0.036 (0.004) 0.031 (0.003)

Deviance information criterion (DIC) 918,336.468 867,208.171 771,270.738

Intraclass Correlation (ICC) 0.012 0.011 0.009

β, log-odds; S.E., standard errors.

(5)

premium quintile groups than in the Medicaid recipient group, while participation rates were higher in the third, fourth, and fifth income quintiles than in the Medicaid recipient group. Par- ticipation rates for subjects with severe disabilities (grades 1 and 2) were lower than for subjects without disabilities, while the participation rate was higher for the mild disability groups.

The proportion of agriculture, forestry, and fishery workers in an area was negatively associated with participation rate (β, -0.005; SE, 0.002). In addition, a higher CDI was significantly as- sociated with a higher participation rate (β, 0.001; SE < 0.0001).

The number of screening centers per 1,000 inhabitants did not significantly influence participation in the screening program (β, 0.493; SE, 0.404). ICCs were low in all models, and when we introduced area-level variables into Model 1, 18% of ICCs de- creased.

When we tested the random slope model of area-level fac- tors, the result was not significant (data not shown). When we tested interaction between different level factors, there was a significant inverse interaction between health insurance pre- mium and the proportion of agriculture, forestry, and fishery

workers in an area (Table 4). The CDI had a positive interaction with health insurance premium, but the number of screening centers per 1,000 inhabitants had no significant interaction with the premium level (Table 4).

Multilevel analysis in the 66-yr-old group

Table 5 shows the results of the multilevel analysis, which in- cluded individual and area-level variables in a hierarchical man- ner. There was significant variance in participation rates across the area in the null model (Fig. 3). Individual-level variables ex- hibited similar effects on participation in screening before and after controlling for area-level variables. The exception was that there was less participation by the Medicaid recipient group in the screening program than all income quintiles of the insured self-employed groups.

The proportion of agriculture, forestry, and fishery workers was positively related with the participation rate (β, 0.022; SE, 0.002). This contradicted the results obtained for the 40-yr-old group. The CDI was significantly positively related to screening program participation (β, 0.001; SE < 0.0001). The number of

Table 4. Interactions between health insurance premium and area-level factors in both age groups

Age group Health insurance premium

Proportion of agriculture,

forestry and fishery workers (%) Composite Deprivation

Index score Numbers of the screening center per 1,000 people

β (S.E) β (S.E) β (S.E)

40 yr old group Medicaid

1st quintile (lowest) 2nd quintile 3rd quintile 4th quintile 5th quintile (highest)

Reference -0.025 (0.002) -0.021 (0.002) -0.019 (0.002) -0.022 (0.002) -0.027 (0.002)

Reference 0.002 (0) 0.002 (0) 0.002 (0) 0.002 (0) 0.002 (0)

Reference -0.248 (0.389)

0.359 (0.385) 0.414 (0.383) 0.496 (0.382) 0.435 (0.389)

66 yr old group Medicaid

1st quintile (lowest) 2nd quintile 3rd quintile 4th quintile 5th quintile (highest)

Reference -0.005 (0.002)

0.001 (0.002) 0 (0.002) -0.005 (0.002) -0.021 (0.002)

Reference 0.001 (0) 0.001 (0) 0 (0) 0.001 (0) 0.001 (0)

Reference 1.034 (0.418) 1.195 (0.431) 2.317 (0.426) 1.860 (0.426) 1.025 (0.408) β, log-odds; S.E., standard errors.

Fig. 2. Regional distribution of the participation rate in the 40-yr-old group, null model.

Estimated relative participation rates were illustrated with 95% confidential intervals.

Relative participation rate

Area

0 70 140 210 280

0.9 0.7 0.4 0.2 0.0 -0.2 -0.4 -0.7 -0.9

Fig. 3. Regional distribution of the participation rate in the 66-yr-old group, null model.

Estimated relative participation rates were illustrated with 95% confidential intervals.

Relative participation rate

Area

0 70 140 210 280

1.0 0.8 0.5 0.3 0.0 -0.3 -0.5 -0.8 -1.0

(6)

screening centers per 1,000 inhabitants had no significant effect on the participation rate (β, 0.578; SE, 0.362). Although ICCs were still low, they were higher than that of the 40-yr-old group in all the models. When we compared Model 2 with Model 1, there was a 40% decrease in the ICC in Model 2.

Although we tested the random slope model of area-level fac- tors, there was no significant effect, as was observed in the 40-yr- old group (data not shown). When we tested the interaction be- tween inter-level factors, the results were similar to those of the 40-yr-old group. Health insurance premium level and the pro- portion of agriculture, forestry, and fishery workers in an area exhibited an inverse interaction (Table 4). There was a positive interaction between CDI score and the health insurance premi- um (Table 4). There was no significant interaction between the number of screening centers per 1,000 inhabitants and health insurance premium (Table 4).

DISCUSSION

To our knowledge, this is the first study investigating the indi- vidual and area-level determinants of participation in the na- tional general health screening program in Korea with a multi- level analysis approach. We determined that regional charac- teristics had significant independent effects on participation in the national screening program. We also compared all Medic-

aid recipient subjects with insured subjects. To date, no study has used the nationwide data of Medicaid recipients to investi- gate their behaviors related health screening.

There were significant regional variations in all the models of both age groups. When we adjusted individual-level factors, ICCs in the 40 and 66-yr-old groups were 0.011 and 0.025, re- spectively. This means that the inter-area variance was smaller for the 40-yr-old group and that individual characteristics had a larger influence in this group. When we adjusted area-level fac- tors additionally, the ICC decrease was more prominent in the 66-yr-old group. This indicates that a larger part of the regional variance can be explained by the area-level factors used in the 66-yr-old group. Consequently, we can expect the 66-yr-old group to be influenced more by regional characteristics than the 40-yr-old group.

As we had hypothesized, level of urbanization had a signifi- cant effect on participation in the national screening program.

However, the effect was different with respect to age. In the 40- yr-old group, the participation rates of residents in urban areas were higher than those in rural areas were. In contrast, in the 66-yr-old group, the participation rates of subjects in rural areas were higher than those in urban areas were. The relationship between urbanization and participation in health screening programs is still controversial. According to previous Korean and Japanese studies (8, 17, 25), the participation rates of sub- Table 5. Multilevel analysis results of the 66-yr-old group

Response Null Model Model 1 Model 2

β (S.E) β (S.E) β (S.E)

Fixed Part

Constant 0.134 (0.017) -0.375 (0.025) -0.688 (0.051)

2009 (vs 2008) 0.187 (0.008) 0.178 (0.009)

Female (vs male) 0.142 (0.008) 0.143 (0.008)

Health Insurance Premium Medicaid Recipient 1st quintile 2nd quintile 3rd quintile 4th quintile 5th quintile

Reference 0.119 (0.016) 0.387 (0.016) 0.512 (0.015) 0.591 (0.016) 0.712 (0.015)

Reference 0.103 (0.018) 0.372 (0.018) 0.491 (0.018) 0.581 (0.018) 0.693 (0.017) Disability Grade

No-disabled 1st grade 2nd grade 3rd grade 4th grade 5th grade 6th grade

Reference -1.250 (0.042) -0.713 (0.026) -0.252 (0.024) 0.095 (0.025) 0.165 (0.021) 0.186 (0.022)

Reference -1.264 (0.043) -0.714 (0.027) -0.261 (0.025) 0.088 (0.025) 0.159 (0.023) 0.186 (0.023)

The proportion of agriculture, forestry and fishery workers (%) 0.022 (0.002)

Composite Deprivation Index Score 0.001 (< 0.0001)

Numbers of the screening center per 1,000 inhabitants 0.578 (0.362)

Random Part : Area Level

Variance (S.E) 0.064 (0.006) 0.086 (0.008) 0.050 (0.005)

Deviance information criterion (DIC) 424,199.224 382,898.537 350,913.859

Intraclass Correlation (ICC) 0.019 0.025 0.015

β, log-odds; S.E., standard errors.

(7)

jects in urban or metropolitan areas in national health screen- ing programs were lower. In contrast, some studies in other coun- tries (12, 26) reported lower participation rates in rural areas.

The role of mobile health screening services may help explain our results. In Korea, according to the Framework Act in Health Examination Law, only agencies licensed by the Minister of Health and Welfare can provide mobile health screening ser- vices outside medical institutions. Mobile health screening ser- vices usually target mountainous and rural areas with poor ac- cessibility to health screening centers. With their instruments, examination staffs access the target location via a specialized bus or container car and conduct general health checkups (phys- ical examination, blood test, urine test, and chest radiography) and cancer screenings (stomach, breast, colon, liver, and cervi- cal cancer). Although there has been some criticism about the low quality of such screening services and quality controls, mo- bile health screening services have steadily been extended to rural municipal districts. In our study, the areas with high partic- ipation rates were concentrated in the mountainous and rural areas (Fig. 4, 5). This was more prominent in the 66-yr-old group.

As most subjects in this group are retired, they may encounter fewer time constraints to utilizing the screening service than those in the 40-yr-old group, who are still actively employed.

This could be responsible for the differing effect of mobile health screening services in both age groups.

In this study, participation rates were significantly higher in deprived areas. This is contrary to previous findings reporting a negative effect of regional deprivation in health screening pro- grams (12, 13, 20, 27). The national health screening program in Korea is a very unique program targeting various diseases and is executed nationwide with political support. Our findings in-

Participation rate (%)

Moran’s I = 0.2388

11.8-27.2 27.2-29.9 29.9-32.2 32.2-35.2 35.2-41.5

-6.0-4.0-2.0 -4.0 -2.0 0.0 2.0 4.0

0.0 2.0 4.0 6.0

A B

Fig. 4. Regional distribution of the participation rate (A) and spatial correlation of the regional participation rate. Values range from -1 (indicating perfect dispersion) to +1 (perfect correlation) (B) in the 40-yr-old group.

Participation rate (%)

Moran’s I = 0.5432

38.3-47.1 47.1-51.9 51.9-55.4 55.4-59.2 59.2-69.1

-3.0-2.0-1.0 -2.0 -1.0 0.0 1.0 2.0

0.0 1.0 2.0 3.0

A B

Fig. 5. Regional distribution of the participation rate (A) and spatial correlation of the regional participation rate. Values range from -1 (indicating perfect dispersion) to +1 (perfect correlation) (B) in the 66-yr-old group.

dicate that interventions and political support targeting deprived areas in the national health screening program, such as through mobile health screening services, advertisements, and remind- er services, may be effective in improving participation rates.

However, the participation rate of the low health insurance pre- mium group was also low at the individual level. When we test- ed interaction between individual-level insurance premiums and regional deprivation, we found that the gap of participation rate according to income level widened as regional deprivation increased in both age groups. This implies that interventions targeting deprived subjects are not sufficiently effective or that the effect of those interventions is not properly delivered to low- income groups. We suspect that the effect of interventions for improving participation mostly reach the higher premium level groups first in deprived areas. Hence, although the participa- tion rate of the deprived area might be higher, there would con- tinue to be inequality at the individual level.

Contrary to our hypothesis, the number of screening centers per 1,000 inhabitants had no significant effect on participation in the screening program. Pornet et al. (13) reported no relation- ship between the number of screening centers and participa- tion in cancer screening in France. However, other studies (28, 29) have suggested that the density of screening facilities is im- portant in cancer screening participation. Our result demon- strates that a mere numeric increase of medical supplies in the national health screening program does not lead to an increase in the participation rate. We believe that mobile health screen- ing services may play an important role in areas with few screen- ing centers. In Korea, there are more screening centers in urban areas than in rural areas. Mobile health screening services alle- viate the shortage of centers and improve accessibility. Hence,

(8)

this leads to a weakening of the effect of the number of screen- ing centers. We can also expect several screening centers to con- duct most of the national health screening programs in an area intensively, and that numerous screening centers are not re- quired.

We also determined that all individual-level variables (i.e., age, sex, health insurance type, insurance premium, and dis- ability grade) had significant impacts on participation even af- ter adjusting for area-level variables. Previous Korean studies (8, 14, 18, 26) analyzed these variables at a single level using multi- variate regression models. Such models can under/overestimate the effect of regional characteristics. Multilevel analysis allowed us to simultaneously investigate and explain the effects of indi- vidual and area-level variables. When we carried out multilevel analysis, these individual-level variables were consistent after adjusting for area-level variables.

Although income level was positively associated with partici- pation, which is consistent with the results of previous studies (14, 25, 29, 30), the participation rates in the first and second income quintiles in the 40-yr-old group were lower than that of Medicaid recipients. In the National Screening Program for Transitional Ages, all health screening services are offered free of charge and there is almost no economic barrier. However, since screening programs are conducted during the day and on weekdays, working people may face time constraints, which may hinder their participation in the screening program. Although most in the 66-yr-old group were retirees, most in the 40-yr-old group had jobs and were actively employed. Unlike the insured group, 40-yr-old Medicaid recipients did not have regular jobs and had relatively fewer time constraints on participation in screening programs. Differing time constraints based on age, insurance type, and income level might play an important role in the difference in participation rates. Thus, the 66-yr-old group and Medicaid recipients might participate more in health screen- ings than the 40-yr-old insured group. A subsequent study us- ing variables on time constraints and their burden would be necessary to evaluate the role of time constraints.

Our study has some limitations. First, although we used the complete data of the 40 and 66-yr-old screening program sub- jects, their results cannot be generalized to the entire popula- tion. However, if we focus our findings on the National Screen- ing Program for Transitional Ages, our restricted sample is en- tirely acceptable. Second, we could not include dependents of the insured employee group in our study. They are not obliged to participate in the national screening program, and their health behaviors may differ from our study population. Future studies should include dependents of the insured employee group and evaluate their characteristics. Third, this study did not include private health screening programs. Private health screening programs are common in Korea. The participants of private health screening programs may not participate in national health

screening programs. It was impossible to determine the actual participation rates of private health screening programs from our data. However, participation in private health screening programs is more common among higher-income groups and those with health insurance. Therefore, the trends of participa- tion rates according to income level would not change signifi- cantly even if we had included the data of private screening pro- grams. Fourth, when we adjusted for all individual and area-lev- el variables, significant unexplained regional variance in partic- ipation rates remained. This means that there are other factors affecting participation in national health screening programs. If we were able to include variables about attitudes towards screen- ing programs, the quality of screening centers, and barriers to participation, we could elucidate this variance in greater detail.

Finally, data regarding individual residence duration were un- available. It is possible that regional characteristics completely affect individuals over time.

In conclusion, our study demonstrates that regional charac- teristics have independently significant effects on individual participation in national health screening programs. They also have different interactions with individual characteristics such as age and income level. To increase participation rates in na- tional health screening programs, tailored approaches that con- sider the characteristics of the target population and region si- multaneously are required.

ACKNOWLEDGMENTS

We thank the staffs of the Korean National Health Insurance Corporation for their thoughtful assistance. The authors declare no conflict of interests.

REFERENCES

1. Lee WC, Lee SY. National health screening program of Korea. J Korean Med Assoc 2010; 53: 363-70.

2. Park JH, Lee JS, Lee JY, Hong JY, Kim SY, Kim SO, Cho BH, Kim YI, Shin Y, Kim Y. Factors affecting national health insurance mass screening par- ticipation in the disabled. J Prev Med Public Health 2006; 39: 511-9.

3. Park JH, Lee JS, Lee JY, Gwack J, Park JH, Kim YI, Kim Y. Disparities be- tween persons with and without disabilities in their participation rates in mass screening. Eur J Public Health 2009; 19: 85-90.

4. Sung NY, Park EC, Shin HR, Choi KS. Participation rate and related so- cio-demographic factors in the national cancer screening program. J Prev Med Public Health 2005; 38: 93-100.

5. Kwak MS, Park EC, Bang JY, Sung NY, Lee JY, Choi KS. Factors associat- ed with cancer screening participation, Korea. J Prev Med Public Health 2005; 38: 473-81.

6. Kim RB, Park KS, Hong DY, Lee CH, Kim JR. Factors associated with cancer screening intention in eligible persons for national cancer screen- ing program. J Prev Med Public Health 2010; 43: 62-72.

7. Rutter DR. Attendance and reattendance for breast cancer screening: a

(9)

prospective 3-year test of the theory of planned behaviour. Br J Health Psychol 2000; 5: 1-13.

8. Hahm MI, Choi KS, Kye SY, Kwak MS, Park EC. Factors influencing the intention to have stomach cancer screening. J Prev Med Public Health 2007; 40: 205-12.

9. Kye SY, Han EO, Park KH. Predictors of regular gastric cancer screening among Koreans. Asian Pac J Cancer Prev 2010; 11: 1315-20.

10. Robert SA. Community-level socioeconomic status effects on adult health.

J Health Soc Behav 1998; 39: 18-37.

11. Pruitt SL, Shim MJ, Mullen PD, Vernon SW, Amick BC 3rd. Association of area socioeconomic status and breast, cervical, and colorectal cancer screening: a systematic review. Cancer Epidemiol Biomarkers Prev 2009;

18: 2579-99.

12. Siahpush M, Singh GK. Sociodemographic variations in breast cancer screening behavior among Australian women: results from the 1995 Na- tional Health Survey. Prev Med 2002; 35: 174-80.

13. Pornet C, Dejardin O, Morlais F, Bouvier V, Launoy G. Socioeconomic determinants for compliance to colorectal cancer screening: a multilevel analysis. J Epidemiol Community Health 2010; 64: 318-24.

14. Jang SN, Cho SI, Hwang SS, Jung-Choi K, Im SY, Lee JA, Kim MK. Trend of socioeconomic inequality in participation in cervical cancer screening among Korean women. J Prev Med Public Health 2007; 40: 505-11.

15. Lee K, Lim HT, Park SM. Factors associated with use of breast cancer screening services by women aged >or= 40 years in Korea: the third Korea National Health and Nutrition Examination Survey 2005 (KNHANES III). BMC Cancer 2010; 10: 144.

16. Roh WN, Lee WC, Kim YB, Park YM, Lee HJ, Meng KH. An analysis on the factors associated with cancer screening in a city. Korean J Epidemiol 1999; 21: 81-92.

17. Fukuda Y, Nakamura K, Takano T. Reduced likelihood of cancer screen- ing among women in urban areas and with low socio-economic status:

a multilevel analysis in Japan. Public Health 2005; 119: 875-84.

18. Chun EJ, Jang SN, Cho SI, Cho Y, Moon OR. Disparities in participation in health examination by socio-economic position among adult Seoul residents. J Prev Med Public Health 2007; 40: 345-50.

19. Fukuda Y, Nakamura K, Takano T. Accumulation of health risk behav- iours is associated with lower socioeconomic status and women’s urban

residence: a multilevel analysis in Japan. BMC Public Health 2005; 5: 53.

20. Dailey AB, Brumback BA, Livingston MD, Jones BA, Curbow BA, Xu X.

Area-level socioeconomic position and repeat mammography screening use: results from the 2005 National Health Interview Survey. Cancer Epi- demiol Biomarkers Prev 2011; 20: 2331-44.

21. Kim HS, Shin DW, Lee WC, Kim YT, Cho B. National screening program for transitional ages in Korea: a new screening for strengthening primary prevention and follow-up care. J Korean Med Sci 2012; 27: S70-5.

22. The National Assembly of the Republic of Korea. Disabled persons wel- fare act. Available at http://likms.assembly.go.kr [accessed on 4 Novem- ber 2012].

23. Shin H, Lee S, Chu JM. Development of composite deprivation index for Korea: the correlation with standardized mortality ratio. J Prev Med Pub- lic Health 2009; 42: 392-402.

24. Chu JM, Park JI, Cho MR, Choi SJ, Im JH, Shin HS. Environmental poli- cy for low-income population in urban area. No RE-02S. Seoul: Korea Environment Institute, 2007, p106-9.

25. Kwon YM, Lim HT, Lee K, Cho BL, Park MS, Son KY, Park SM. Factors associated with use of gastric cancer screening services in Korea. World J Gastroenterol 2009; 15: 3653-9.

26. Maxwell CJ, Bancej CM, Snider J. Predictors of mammography use among Canadian women aged 50-69: findings from the 1996/97 National Pop- ulation Health Survey. CMAJ 2001; 164: 329-34.

27. Schootman M, Jeffe DB, Baker EA, Walker MS. Effect of area poverty rate on cancer screening across US communities. J Epidemiol Commu- nity Health 2006; 60: 202-7.

28. Meersman SC, Breen N, Pickle LW, Meissner HI, Simon P. Access to mam- mography screening in a large urban population: a multi-level analysis.

Cancer Causes Control 2009; 20: 1469-82.

29. Coughlin SS, Leadbetter S, Richards T, Sabatino SA. Contextual analysis of breast and cervical cancer screening and factors associated with health care access among United States women, 2002. Soc Sci Med 2008; 66:

260-75.

30. Kim YB, Ro WO, Lee WC, Park YM, Meng KH. The influence factors on cervical and breast cancers screening behavior of women in a city. J Ko- rean Soc Health Educ Promot 2000; 17: 155-70.

수치

Table 1. Description on the Composite Deprivation Index
Table 2 lists the baseline characteristics of the study population.
Fig. 3. Regional distribution of the participation rate in the 66-yr-old group, null model
Fig. 4. Regional distribution of the participation rate (A) and spatial correlation of the  regional participation rate

참조

관련 문서

Factors associated with hypertension control in Korean adults: The fifth Korea National Health and Nutrition Examination Survey (KNHANES V-2).. Association of stress

The related factors with clustering healthy behaviors by multinominal logistic regression analysis were higher for “females, the elderly, people with higher level of

Objective: The purpose of this study is to investigate the relationship between depression and evening meals for adult women by using the National Health and Nutrition Survey

Prevalence of Undiagnosed Diabetes and Related Factors in Korean Postmenopausal Women:.. The 2011-2012 Korean National Health and

Objective: This study was conducted to identify the association between vitamin D and Sarcopenia among all adults in Korea using data from the National Health and

As a result of multiple regression analysis, the statistically associated factors with burnout were social support, GDS, marital state, self­rated health

We find the correlation between marital status and current smoking, including interaction with marriage, occupation level, and Year.. Methods : We used the Korea National

Second, there was no statistically significant difference in the sub-factors of participation satisfaction according to age, career, occupation,